We charge a fixed plan. Token $ you save stay yours.
Hobbyists and evaluation
Indie founders in production
Typical savings $2,000–$5,000 / month
Startups and agencies
Typical savings $20,000–$50,000 / month
Compliance and control
Volume is requests per 30 days — not tokens. The plan fee is a flat SaaS fee for the proxy and dashboard; provider spend stays on your own BYOK key, and output tokens usually cost 3–5× more per million than input.
Live today: OpenAI + Anthropic proxy, streaming, BYOK, exact-match cache, heuristic routing, code_ast compression, failover, budget headers, request/key/user limits, dashboard spend and savings. Embedding-based semantic cache, email/Slack alerts, evals, PII redaction, SSO and compliance controls are included in their tier as they ship — they are not shipping yet and not billed separately.
Hosted platform keys, when used, run on gpt-4o-mini under a separate cost cap. That cap is an infrastructure limit, not the marketing request gate.
| Capability | Free | Pro | Team | Enterprise |
|---|---|---|---|---|
| Volume & limits | ||||
| Requests / 30 days | 10,000 | 500,000 | 5,000,000 | Negotiated |
| API keys | 1 | 5 | Unlimited | Unlimited |
| Users | 1 | 1 | 10 | Unlimited |
| Cache TTL | 1h | up to 24h | up to 7d | Custom |
| Log retention | 7 days | 30 days | 90 days | Custom |
| Core proxy | ||||
| OpenAI + Anthropic | Yes | Yes | Yes | Yes |
| Streaming | Yes | Yes | Yes | Yes |
| Exact-match cache | Yes (shared global) | Yes | Yes | Yes |
| Semantic (embedding) cacheas it ships | — | Yes | Yes | Yes |
| Heuristic routing | Auto only | Manual override | Custom policies | Custom policies |
| code_ast compression | Yes | Yes | Yes | Yes |
| Pro controls | ||||
| Manual model override | — | Yes | Yes | Yes |
| Failover routing | — | Yes | Yes | Yes |
| Hard spend caps | — | Yes | Yes | Yes |
| Request tracing | — | Yes | Yes | Yes |
| Email / Slack alertsas it ships | — | Yes | Yes | Yes |
| Team collaboration | ||||
| Per-project breakdown | — | — | Yes | Yes |
| Per-user caps | — | — | Yes | Yes |
| Per-endpoint cache TTLas it ships | — | — | Yes | Yes |
| LangChain / LlamaIndexas it ships | — | — | Yes | Yes |
| A/B + shadow trafficas it ships | — | — | Yes | Yes |
| Eval suiteas it ships | — | — | Yes | Yes |
| PII redactionas it ships | — | — | Yes | Yes |
| Enterprise & compliance | ||||
| SSOas it ships | — | — | — | Yes |
| RBACas it ships | — | — | — | Yes |
| Audit logsas it ships | — | — | — | Yes |
| SOC 2as it ships | — | — | — | Yes |
| On-prem / VPCas it ships | — | — | — | Yes |
| Data residencyas it ships | — | — | — | Yes |
| White-labelas it ships | — | — | — | Yes |
| Service | ||||
| Uptime SLA | None | None | 99.5% | 99.9% |
| Support | Community | Email 48h | Live chat 4h | Dedicated CSM |
Rows marked “as it ships” are included in that tier when the capability ships. Free hits 10,000 requests or needs a second key → upgrade to Pro. Need team visibility → Team. Need SSO or compliance → contact sales.
Plan fee only. The $99 Pro / $499 Team fee is a flat SaaS fee for the compression proxy and dashboard. Upstream usage is billed by your provider on your BYOK key, and output tokens usually cost 3–5× more per million than input tokens. Set hard limits in your provider dashboard so users cannot overcharge the account.
Requests per 30 days — not tokens. Free is 10,000, Pro is 500,000, Team is 5,000,000, Enterprise is negotiated.
OpenAI + Anthropic proxying, streaming, BYOK, exact-match cache, heuristic routing, code_ast compression, failover, budget headers, request/key/user limits, and the spend/savings dashboard. Embedding-based semantic cache, email/Slack alerts, evals, PII redaction, SSO and compliance controls are included in their tier as they ship — not shipping today.
Free hits 10k requests or needs a second key → Pro. Need team visibility, per-project breakdown or per-user caps → Team. Need SSO, RBAC or compliance → contact sales.
Modes control aggression. Start lossless; raise it only if you want more savings. Typical result is 30–60% fewer input tokens — not a guaranteed cut to your total provider bill.
Anything with a custom OpenAI base URL. Not Copilot or Cursor Tab.
Free tier. No card.